costs. It was assumed that the average salary per teacher at the different levels of education would increase at the same rate as per capita GNP,* as would the average capital cost per new pupil place. Total GNP was assumed to increase at the same rate in all situations (6 percent per annum), which meant that per capita GNP increased more rapidly in the declining fertility projections. Two assumptions were made about the share of teacher salaries in total recurrent expenditure: one, that this would remain at its 1960 level of 80 percent; the other, that it would fall to 65 percent in 1990, implying an improvement in the quality of education as teachers are better supplied with textbooks, teaching aids, etc.
According to these assumptions, total costs increase spectacularly in all projections, but then so does GNP. The important consideration is whether educational costs rise in relation to GNP. Evidence on this is given in Figure 4, which also shows the percentage of GNP saved in the given years and according to the given targets if fertility declines according to the pattern of Projection III. Clearly, the main factor governing trends in educational costs is trends in enrollment rates. Where these are held constant, the share of GNP required for education rises only slightly especially if fertility declines; where enrollment rates rise as postulated, this share grows manyfold, even if fertility declines rapidly. If pupil/teacher ratios are lowered as well, and the non-teacher salary component of recurrent costs raised, education's share of the GNP climbs to 6 percent and above by 1985, even if fertility declines rapidly. This figure does not include the costs of higher education, including teacher training, which will also be increasing and might well account for a further 1 or 2 percent of GNP by the end of the projection period. At the present time 8 percent of GNP is as much as any country in the world spends on education, and although it would not be impossible for Pakistan to reach this level from the 1.7 percent of GNP actually spent on education in 1965, this would require a very basic reordering of priorities.
A rapid decline in fertility after 1965 will have very little effect on costs for about 10 years, but Figure 4 demonstrates that it will subsequently be of substantial benefit in the struggle to upgrade the education system, and its effect will grow cumulatively as time goes on. By 1990, for example, declining fertility would result in a saving of 1.6 percent of the GNP each year in reaching the goal of increased enrollment rates and improved pupil/teacher ratios: a figure almost identical to the entire share of the GNP going to education in 1965.
Moreover, the relative contribution of a decline in fertility to lowering costs is of the same general order of magnitude if attempts are made to raise
*The validity of this assumption will depend largely on the qualification level of new teachers entering the teaching force. Pay scales are linked both to length of service and qualification. When the teaching force is increased sharply, as in the projection with increasing enrollment rates, average length of service of the teaching force will fall. Thisary structure. In developing countries level of qualifications is normally given far more weight.and